A few years back,I recall an item written by a member of the dPreview staff (Barnaby Britton maybe?) that suggested that lens testing folks had stumbled onto a type of what I would classify as a kind of
digital reciprocity failure, in which digital camera sensors seemed to break the reciprocal arrangement in which one stop wider and aperture resulted in one full EV more exposure--once lens apertures climbed above (
wider hole than) f/1.4 I believe it was. This was as I recall, based on test results using some of the fairly new ultra-speed lenses which have recently proliferated in the mirrorless realm--lenses like the new-ish f/0.95 Cosina-made lenses, for example.
It's been a while, but as I recall, the piece seemed to suggest that digital sensors were just not delivering fully reciprocal exposure levels as the lens apertures grew wider and wider...that there was in effect, what one could describe as a loss of effective ISO speed. I sure as heck don't have any uber-speed lenses to test this out, and I almost never shoot any of my lenses at wider than f/2.8, even though I have a number of f/2 and f/1.8 and f/1.4 lenses...
******Okay....I found one thread about this....
Light loss on current CMOS sensors at big apertures: Open Talk Forum: Digital Photography Review
Light Loss on current CMOS Sensors at Big Apertures. The simplified concept would be that as pixel wells have grown smaller and smaller as sensor density has elevated, that light coming in at oblique angles has a tougher time at "filling" the so-called pixel wells.
dPreview's Andy Westlake chimes in in the above thread with test results from a Leica 50mm f/0.95 Leica Noctilux lens that shows that the depth of field blurring does NOT follow the proper linear relationship at apertures wider than f/1.4. Again...this discussion was a while back, and it might be that smaller-sized sensors and super-speed lenses don not deliver all of their theoretical performance.
Speaking of
aperture and ISO and the relationship: I think a much bigger,much more real-world issue might be effective light transmission, aka
T-stop, and how that relates to the number of lens elements, and how many air-to-glass surfaces there are in a lens, and how well-designed the anti-reflection coating is for the lens design. SOME lenses, like say a certain Nikon 70-200 f/2.8 zoom with 21 elements in 15 groups, seem to me to have an actual T-stop that's well below what say, my 135 D.C. lens achieves...the 135 DC is my "brightest" lens, with a mere 7 lens elements in 6 groups.
T-stop is a big deal to cinematographers. Their lenses are marked in T-stops.